ARI-RR- 1029

INFRARED FIBER OPTIC DIAGNOSTIC OBSERVATIONSOF SOLID PROPELLANT COMBUSTION

Final Report

SI ELECTE

SFE SB 2 5 1994

APrepared by

J. Wormhoudt, P. L. Kebabian, and C.E. KolbCenter for Chemical and Environmental Physics

Aerodyne Research, Inc., 45 Manning Road, Billerica, MA 0 1821

Prepared forU.S. Army Research Office

Under Contract DAAL03-9 I -C-00X)8

October 1993

Best Available Copy

APPROVED FOR PUBLIC RELEASE,DISTRIBUTION UNLIMITED

REPORT DOCUMENTATION PAGE OMB No.0-0,

,;at"O"N ane 'omlww• IN cats "@pgo" 8"d foon i" &no N M@e o#"OIPon Of ,-9•_Svoý S& a O €M-""tI• t",% curgaP "tImalt Of attY otnot 0%0,M~ or " I

ýoflifiJof ot ,.,WtondtonA."(rU4'fl@ "9ml onso ,euct"~ inn .,oue to vv~owo -.m.~a~un#" ei e,~. .,'@"ofte ter~o.t otIjeawao ,2*.n ""Me _ • w 10 1204 -ot1.<lton. 2 fl2,:;24302 m to &1" Oflne ^* M$ iffmn• t an4 B.,G-q¶ *.eo .o AE~ueon $ofdtO .O44lUI.WlAft@O. 3C 20SO)

1. AGENCY USE ONLY (Leavt o4ano. ) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED

30 Novemoer '1 -93 4 . i., _enort. 1, 21/91-9/30/93

4. TITLE AND SUBTITLE 5. FUNDING NUMBERS-nfrared -ioer "••i ia~nosr ic bservations

`f oiid P-ropellant .Nu~o

D4&63-9,'-C_.o6L?6. AUTHOR(S)

', ;ormnoudt, PL. :,ebabian, and C.E. Kolb

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATIONREPORT NUMBE[R

Aerodvne Research, Inc.

-ý5 Manning Road ARI-RR-1 029Billerica, MA 01821

9. SPONSORING / MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING / MONITORINGAGENCY REPORT NUMBER

U. S. Army Research OfficeP. 0. Box 12211Research Triangle Park, NC 27709-2211 Zt/e y/0)./-.I--

11. SUPPLEMENTARY NOTESThe viev, opinions and/or findings contained in this report are those of theauthor(s) and should not be construed as an official Department of the Armyposition, policy, or decision, unless so desi=nated by other documentation.

IZa. DISTRIBUTION /AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE

Approved for public release; distribution unlimited.

13. ABSTRACT (Maximum 200 wOres)We report on work to develop and demonstrate a diagnostic technique using

.n.rared fiber optics to probe the decompositicn processes in burning gun7ropellant strands. The experiments reported here involve measuring the absorpticnacross an open gap between two embedded fibers as it fills with gaseousiecomposition products. Spectroscopic detection is achieved using pairs ofbandpass filters. The absorption record can be correlated with readings fromembedded thermocouples and with a high resolution video recording of the burn. Wehave observed N2 0, a decomposition product of RDX, evolving into the observationvolume. The N2 0 appearance is often an abrupt event. Coupled with evidence thatit can occur while the observation region is relatively cool and far from theburning surface (observed in cases involving burning at 6 atm pressure), theseobservations suggest hypotheses concerning the physical processes in the condensedphases of this propellant, hypotheses including the development of pressurizedvoids or matrix-trapped bubbles which can release decomposition gases into theunburnt propellant through suddenly opening cracks. The appearance of N2 0 after-he temperature record suggests that melting and decomposition of RDX have begun(observed in 1 atm cases) implies that N2 0 formation is indeed not an initial stepin RDX decomposition under our conditions.

14. SUBJECT TERMS IS. NUMBER OF PAGESsolid propellant combustion 41

infrazc:d fibcr opLic absorption diagnostic 16. PRKI CODE

17. SECURITY CLASSIFICATION 1I. SECURITY CLASSIFICATION IS. SECURITY CLASSIFICATION 20. UMITATION OF ABSTRACT

OF REPORT OF ABSTRACT

UNCLASSIFIED UNCLASSIFIED UNCLASSIFIED UL

NSN 7S40-01.i0SSOO Standard Form 298 (Rev 2-39)ofw9 w -Mn Us am'~smn&W

TABLE OF CONTENTS

A B ST R A C T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.0 INTR O D U CT IO N ............................................. 1

2.(0 EXPERIMENTAL TECHNIQUES .................................. 2

3.0 EXPERIMENTAL ANALYSIS . .................................... 73.1 Nitrous O xide Q uantification . ...................................... 73.2 Flame Front Position M easurement .................................. 73.3 Propellant Temperatures Measurements ................................ 8

4.0 EXPERIMENTAL RESULTS ..................................... 94. I Exam ple Data Traces .. ........................................... 94.2 N um erical Results . ............................................. 14

5.0 THERMAL W AVE ANALYSIS .................................... 18

6.0 DISCUSSIO N . ............................................... 276.! Correlation of Observed Events with Temperature History .................. 276.2 Correlation of Our N20 Observations with Chemical Mechanisms of RDX

Decomposition ................................................. 32

7.0 SU M M A R Y . ................................................. 36

X.0 REFERENC ES . ............................................... 37

9.(0 LIST OF Pt IBLICATIONS AND TECHNICAL REPORTS ................. 39

1(). LIST OF PARTICIPATING SCIENTIFIC PERSONNEL .................. 40

II. REPORT OF INVENTIONS ...................................... 41

LIST OF FIGURES

Number E"11

I Schematic Drawing of Infrared Fiber Optic Absorption Experiment. 3

2 Cross-Section of Strand Burner. 4

3 Fiber Insertion Procedure. 6

4 Example Observations for XM39 Burning at 6 atm Pressure. (Labelsidentifying the curves will apply to Figures 5 through 9.) 1 1

5 Example Observations for XM39 Burning at I atm Pressure. 12

6 Example Observations for XM39 Burning at I atm Pressure, ShowingSlower N-,O Evolution. 12

7 XM39 Burning at I atm Pressure, Showing Slow Appearance andDisappearance of N20. 13

8 XM39 Burning at 6 atm Pressure, Showing Slow Appearance andDisappearance of N20. 13

9 XM39 Burning at 6 atm Pressure, With Kapton Film Inserted AboveFiber Gap. 14

10 Temperature History of 6 atm Case Shown in Figure 4, Showing Fits toSimple Exponential Expressions (Dotted Lines). 20

I 1 Temperature History of I atm Case Shown in Figure 5, Showing Fits toSimple Exponential Expressions (Dotted Lines). 20

12 Thermocouple Reading versus Flame Front Position of 6 atm Case Shownin Figure 4 (Solid Line). Average Exponential Data Representations(Dashed Lines), and Error Limit Expressions (Dotted Lines). 23

13 Thermocouple Reading versus Flame Front Position of I atm Case Shownin Figure 5. (Line types have same definitions as in Figure 13.) 23

14 (Comparing Simple Exponential Data Representation and Error LimitExpressionsfor 6 atm Cases (Dashed and Dotted Lines) with Equation IExpression Including Temperature-Dependent Thermal Diffusivity andOther Table 3 Parameters (Solid Lines). 24

:5 Comparing Simple Exponential Data Representation and Error LimitExpressions for I aim Cawes (Dashed and Dotted I ines) with Equation IExpression Including Temperature-Dependent Thermal Diffusivity andOther Table 3 Parameters (Solid Lines). 24

iii

LIST OF FIGURES (Continued)

Number Pace

16 Thermal Diffusivitv for XM39: Low Temperature Measurements 14 (SolidLine). Data Point from Strand Burning Observations of Reference 15(With Quote Error Limits. Plotted at Observed Inflection Point Temperatureof 165 'C), and Our Linear oc(T) Expressions From Table 3. Used inFigures 14 and 15 (Dot-Dash and Dashed Lines). 25

17 Thermal Diffusivity for XM39: Low Temperature Measurements 14 (SolidLine), Data Point from Strand Burning Observations15, and Range ofValues whose Equation I Temperature Histories Fall within the ErrorLimit Regions in Figures 14 and 15 (Dots). 26

18 Qualitative Sketch of Typical Thermal Waves and Observed Event Positions. 28

19 Major RDX Decomposition Gas Species from Experiments of Fetherolf andLitzinger,1 I Compared to Adaptation of Observations of Behrens (dots). l 31

20 Major Gas Phase Products of RDX Primary Combustion, from ModelCalculations of Melius. 19.20 31

Ar . ,, .

ByivI :..

.(¾ , . I,..

iv

LIST OF TABLES

Number

I Composition of XM39 Propellant 5

2 Averages and Ranges of Quantities Measured in XM39 Experiments 15

3 Parameters Obtained from Thermocouple Temperature Histories 21

4 N2O Formation at Various Heating Rates 35

V

I1.0 INTRODUCTION

Recently, a study group sponsored by the Army Research Office developed and published

an overall basic research plan to guide advanced research on key nitramine propellant ignition and

combustion issues which must be addressed to allow the systematic exploitation of this exciting

class of energetic materials. I Subsequent workshops organized by the Office of Naval Research

and the JANNAF Combustion Subcommittee on Nitramine Propellants have further refined the

outstanding issues. The work reported here is a response to a need identified by all these groups,

the development of new diagnostic techniques to monitor the progress of decomposition and

ignition kinetics in the condensed phase.

There have been very few direct measurements of condensed phase decomposition in

burning gun propellants. Our technique, motivated by recent advances in electro-optical material.

is the use of embedded optical fibers. We have begun by investigating the use of

infrared-transmitting fibers in a direct absorption experiment. in which a gap between a pair of

embedded fibers may allow spectroscopic detection of decomposition products within a strand of

solid propellant.

Although ours is the first spectroscopic investigation in the solid phase of burning

propellants. it has benefitted from a number of related research programs (a recent comprehensive

review of experimental and theoretical work is found in Reference 2). Infrared absorption has

been used to study the decomposition of propellant ingredients. including the nitramines HMX and

RDXI-7 i, s well as other materials.8 These studies all used Fourier transform infrared

spectrometers to measure the changing absorption of thin films of the materials ats they underwent

rapid heating 3"6 or photolysis by an ultraviolet lamp. 7 In particular, the spectra reported in

Reference 3. in which the absorption bands of N20 and CO2 can be seen as these gases accumulate

in bubbles forming in confined thin films of HMX and RDX. did much to encourage us in our

approach. Other programs providing useful information have included the work by Miller on

embedding fine-wire thermocouples in burning propellant strands,9 the ongoing work at Sandia

National Laboratories involving infrared spectra ofi heated RDX films.I0 and work by Litzinger

and co-workers on mass spectrometer sampling of laser-heated propellant samples. II

2.0) EXPERIMENTAL TECHNIQUES

Our goal was to develop a technique of fast-response spectral infrared fiber probing of

burning propellant strands, and demonstrate it in mechanism studies of condensed phase processes

in nitramine propellant combustion. In early stages of this work, we focused on observingchemical changes in solid propellant by infrared absorption, embedding a pair of fibers separated

by a sliver of propellant. We will not report on that work here, but instead confine our discussion

to a second experimental configuration which has been much more successful. In these

experiments, the gap between fibers (inside the propellant strand) was lengthened to 0.2 to 0.5 cmand left empty. This air-filled cavity or observation volume could then fill with gas-phase

decomposition products as the flame front approached. Observations including when. how fast,and to what levels it was filled could be used to support hypotheses as to the source of the

decomposition gases.

Previous studies. 1-3.11 indicated that nitramine propellant ingredients such as RDX may

decompose into a number of gaseous species including N2O. H2CO. NO1. HCN. H-,O. NO, andCO1 . The fiber length Used here of about I m means the transmission band of the fiber matches

the response curve of the detectors, both extcnding from less than 2 to about 5 pnm. This allows

sensitive detection of N2O iM• the v3 band around 4.7 ptm, as well as CO- in the band around

4.3 pm, H1CO around 3.5 pm and, with a laser source. NO around 5.2 pim. Of these, N2O was

chosen as the first example because, of those species expected to give strong infrared absorptionsigenals with the first version of our apparatus, it was of greatest interest for the definition of

chemical mechanisms of decomposition. Extensions of optical fiber probe techniques to othervisible and infrared wavelengths to quantify other potential decomposition species including NO-

are being explored in a follow-on effort.

A schematic diagram of the experiment reported here is illustrated in Figure 1. Thechopping wheel modulates the light from an infrared source (a globar). and so the transmittedsignal. at approximately 2()() Hz. The light is f1kcused onto a fluoride glass optical fiber (InfraredFiber Systems) of 200 pm nominal diameter. The propellant strand is mounted inside a high

pressure strand burner (shown in Figure 2). The embedded optical fibers pass through thebaseplate of the strand burner. as do leads connecting a variac to a V-shaped segment of nichrome

wire pressed against the top of the strand to provide ignition. In the observations reported here,

the gaps between the source and detector fibers are in the 0.2 to 0.5 cm range. Light collected by

the fiber on the detector side is divided by a 50/50 beamsplitter and passes through a pair of narrowband infrared interference filters before being focused onto InSb infrared detectors.

'odulatino

Fber-E edde ',pr

Propellant Strand _hermal Sourrp

Infrared/re-7Fiber Focusing

LnSynch Signal toComputer Trigger Input

"Beamsolitter

Figure I. Schematc Dr11awing 1Ifae ie picAsrto xeiet

On-Band,O•ff-Bland

FarofIterfP(nc

To Amphlfer

Sand Cornirutpr Ar

Lens

Infrared '

Detector

Figure 1. Schematic Drawing of Inf'rared Fiber Optic Absorption Experiment.

A data acquisition computer reads a signal point from each infrared detector at each on and

olf point of the chopper. Computer data acquisition for each chopper cycle is triggered by the

synchronization signal from the commercial chopper c.ontroller. The signals from the infrared

detectors and pre-amplifiers pass through butftr amplifiers to minimize the effects ol' the analog to

digital conversion. This also allows optimization of the time response of the signal. so the single

data point the computer reads is representative of the entire half-cycle of the chopper.

The strand burner shown in Figure 2 is a copy of a device in use at the U.S. Army

Research Laboratory, Aberdeen Proving Ground, with some modifications to accommodate fiber

optics. The only propellant studied to date has been XM39, composed of 75%7, RDX and 2517c

binder. Its detailed composition is shown in Table I. Experiments were performed at 2 pressures.

I and 6 atm. Although both pressures are far from those encountered in rocket combustion

chambers, let alone in cannon shells, it turns out that the variation in burning rate between the two

pressure regimes is large enough to give qualitatively different observations and hence useful

infokrmation.

P,',P SSURIZATIOfI`EXHAUST PORT

S~OBSERVATION•\ WINDOWS

PROPELLANT

FIBEROPTIC

TRANSMITTED

DETECTORS

SHIELD GAS FLOWSCOMBUSTION GAS FLOW

Fi _ure 2. Cross-Section of Strand Burner.

... 4

Table 1. Composition of XM39 Propellant

,omponent Name Chemnical Formula Mass Fraction Melting Point

RDX (5 micron) C3H6N606 0.76 l1X) MC

Cellulose Acetate C15H2208 O 0.12 215-225 0CButyrate (CAB)

Nitrocellulose (NC) C6H7 .555O(NO2)2.4 5 0.04 160-17() 0C

Acetyl Triethyl C 14 H22,0 8 0.076 (B. P. 132 "C)Citrate (ATC)

Ethyl Centralite (EC) C17H2 ON2 0.004 721 C

XM39 will not burn in an inert atmosphere at I atm, so we used secondary combustion ofthe gaseous decomposition products with air to support the primary combustion of the solidpropellant. A combination of nitrogen addition to the air and a large enough flow velocity around

the strand was used to prevent burning down the side of the strand. At atmospheric pressure, we

used a gas flow through tile central inlet composed of ,235 STP cm 3/s air and 40 STP cm 3/s N-).

for a gas velocitv around the strand of roughly 3.5 cm/s. At 6 atm, the air and N, flows were

increased to 2470 and 825 STP cm 3/s. respectively, changing the velocity of the gas flow around

the strand to around 7 cm/s. Under these gas flow conditions, the strand burns with a roughly flat

melt layer and a slightly concave burning surface. From measurements discussed below, wefound propellant burning rates of about 0.015 cm/s at I atrm. and 0.041 cm/s at 6 atm.

A high-resolution video system Is used to record each burn. The beginning of datacollection is accompanied by a computer signal to the video recorder which starts its time clock.With a second lens. the video camera is also used as a microscope during fiber insertion.

The fibers are inserted between two sections of a propellant strands, cut with a diamond saw,The fiber insertion procedure begins with the formation of a groove in both propellant pieces. Thestrand segment is placed in a hole in a warmed metal block. At the bottom of the hole is a wirestretched across its diameter. The wire thickness is chosen so that heavy pressure on the strand

leaves the proper indentation on its face. Using a knife, grooves for thermocouples (0.0)127 cm

diameter, chromel-alumel) are cut to within approximately 0. I cm of the fiber groove. Then. after

5

illhfibhers are positioned in thle groove. thyare pressed into thle gi OOC U.sing thle top propellant

piece, thle theri-ocouples are inserted. and both fibers and thermIocouples are Secured inside thle"11rand with cvanoacr-vlate adhesive at the entrv points. Figure 3 illtistrates the techni(que. Thle

reais~sern hled strand has at cylindrical cavity or observation volume with a diamecter of aboUt 0.02 cmn

and at lemuth of 0.2 to 0.5 cm., with thermiocouple lunctions onl one or hoth sides of it. at few fibehr

diamecters out~side the wall of the cavity and from ().(X)5 cmi to 0.02 cmn helow the plane containing

thle f iber.

Adhesive Adhesive

PressedGrooves

forFibers Grooves

for SupportThermocouples Strut

Figtire 3. Fiber Insertion Procedure.C

3.0) EXPERIMENTAL ANALYSIS

3.I Nitrous C(-ide Quantifica.ion

The 501(% transmission points of the filter used to detect N2O were at about 2 185 and2238 cm-1. while those of the reference filter were 2093 and 2196 cm-1. Since the half-heightpoints of the N•O absorption band are at around 2184 and 2254 cm- 1, it is clear not only that theN2O filter closely overlaps the N2O band, but also that the reference filter detector will besomewhat sensitive to the presence of N10 in the absorption path. Initially, we simply includedthis sensitivity in the analysis relating the ratio of the two filter intensities to N20 mole fractions.In later experiments a short cell filled with one atmosphere of N20 was placed in front of thereference filter, so that light reaching the reference detector only had wavelengths which are not

absorbed by N20, thus eliminating the sensitivity.

The filter responses were calibrated computationally. using Fourier transform infrared

spectrometer measurements of the bandpass filter transmission curves, and N2O spectralabsorbance values derived from the HITRAN absorption line database1 2 . Although it waks checkedagainst observations made by filling the strand bumer with a range of pressures of N2O while thegap between fibers was still uncovered, the computational calibration is the primary basis for theobserved mole fractions reported here. because it most easily allows us to take into accountchanges in the in-band absorption due to changes in line strengths and widths with temperature.

The transmitted intensity signals are obtained from computer files of the detector outputs bydifferencing each pair of chopper-open and chopper-closed values. These intensity records arethen averaged- an average of '3 pairs of points yields a 30 Hz data record, and for noisier data10 Ul- or 5 Hz averages are used. After the fiber has burned through, the remaining data stream isaveraged to give a zero level which is subtracted from each detector signal before the ratio is Liken.The value of this ratio, relative to its value before ignition, gives a fractional transmission in theN20 band. It is compared with the calibrations described above to directly yield the product of thepathlength and the N20 number density and, with further assumptions, the N2O mole fraction.

3.2 Flame Front Position Measurement

Tile video record of each burn was examined to yield a record of the position of the burningsurface, synchronized with tile computer records of infrared transmission and thermocoupletemperatures. Flame front positions were typically measured at intervals of one second or more

7

with a time uncertainty1 o0 less than the frame period of' 130) s and a position uncertainty of about0 .02-0).(03 cm (the raster spacing for the range of image magnifications used). These position

histories typically exhibit quite steady burn rates. Some burns do have irregularities near the fiber

insertion point. These irregularities may in some cases be due to a change in the bum rate across

the entire face of the propellant strand, but many apparent irregularities are due to difficulties in

seeing the center of the strand, such that the point chosen to represent the burning surface is on a

front edge. An edge region may bum slower or faster as the fiber insertion point is approached,

because of the heat transfer and air flow perturbations of the fibers, thermocouples. adhesive, and

the cut through the strand.

There could be. theretore, a difference of as much as (). 1 cm in some cases between the

position in the transcription of the video record and the actual distance from the flame front in the

center of the strand to the fiber insertion plane. To give a measure of how much these differences

affect ensemble averages of a quantity such as the fiber-to-Ilame-front distance when N20 first

appears in the observation volume, we will report this distance based both on the point-by-point

video record, and on a least squares fit to the entire position history. The two positions agree far

from the fiber insertion point, and if the burning rate of the center of the strand were more regular

than that of the edges. we might expect the linear fit to give somewhat more reliable fiber-to-flame-

front distances.

3.3 Propellant Temperatures Measurements

As noted above, the thermocouples had 0.0 127 cm wire diameters, and the diameter ot the

bead at the junction is more than double this value. However, we are convinced that the major

soutrce oCf tlficci taintv in the thermlocotple readings is not this relatively large size or issues of good

thermal contact with tile propellant or heat transfer out through the thermocouple. but rather the

toccasional) irregularity in the burning surface in the vicinity of the fibers. As discussed above.

the burning surface far from the fibers is relatively perpendicular to the axis of the strand and is

slightly concave. As the flame bums through the cut in the strand and around the adhesive beads

on tile strand surface which secure the fibers and thermocouples. the slope of the burning surface

can become larger, such that the thermocouple bead could be in a warmer or colder region of

propellant than the fiber gap of order 0). I cm away from it. It should also he said, though, that in

outr burns in which two thermocouples were inserted, the two temperature traces agreed well.

x

4.0 EXPERIMENTAL RESULTS

4.1 Example Data Traces

Figures 4 and 5 present examples of the data traces near the time of N20 appearance for thetwo pressure regimes studied (1 and 6 atm). The figures show the two detector intensities (bothnormalized to one at their maximum values), their ratio, the embedded thermocouple temperature,and the two measures of distance between the burning surface and the fiber. In both cases, theN,,O (as indicated by the drop in the ratio of detector intensities) appears and reaches a fairlyconstant level in a time between 0. 1 and 0.2 s. It must be pointed out that the two pressures differin how often this abrupt N2O appearance is observed, since it is seen in 5 of 6 cases at 6 atm.while in 5 of 7 I atm cases the N2 0 appears over times ranging from 0.3 to 2.0 s.

If we assume that the N-2O is at the temperature of the embedded thermocouples (asmentioned above, an assumption with some uncertainty, but we have no better source of gas

temperature information) we can use the known absorption path lengths to calculate N20 molefractions, which are 0.05 for the 6 atm case in Figure 4 and 0. 18 for the I atm case in Figure 5. Ifwe make the conservative assumption that the ranges of thermocouple temperatures at which N20appears (discussed below) represent the uncertainty in the temperature of the N2 0 in theobservation volume, then from our precalculated filter calibration factors we find that at Iatmosphere this translates into a relative uncertainty of less than -0.3 in N20 mole fraction, whileat 6 atm, the relative uncertainties in N2O vary within the range of ± 0.05 to ± 0.20 as the

transmission varies from 0.8 to 0.2.

After its initial abrupt appearance. the N-0 level (and the infrared transmission through theobscrvation volume) is maintained for an average of about 0.4 s in the 6 atm cases and for betweenI and 2 s in the I atm cases (in the I atm cases with more gradual N2O appearance the timebetween first appearance and loss of transmission is also on the order of I s). We will report twotime segments for the abrupt appearance cases, referring to them as fill times and steady statetimes, in spite of the facts that in other cases they are part of a continuous event of N2 0 appearance

in the observation volume, and that N20 absorption after the initial fill is only relatively constant.On this latter point, we can note that in Figure 4 a change in N20 temperature from 1(X) to 140 'C

(following the rising thermocouple reading) would be expected to cause the transmission toincrease from 0.74 to 0.78- in fact, if anything it decreases, suggesting that the N2 0 molefraction is increasing in the observation volume, although at a much slower rate.

9

After some time at a fairly steady N-(O level, a second relatively abrupt event brings the

transmission to zero. It can he seen in this region that the absolute transmission can drop hy an

order of magnitude while the ratio of the two transmissions remains constaut. Therefore. the

experiment is concluded by events whose effects on the transmission have no spectral dependence

over the range of the handpass filters, such as misalignment of the fibers or entry of liquid into the

observation path.

Figures 4 and 5 exemplify a second difference between the two pressure regimes. that the

appearance of the N-0 takes place in different temperature ranges - at under I W) °C in the 6 atm

example. but when the thermocouple reading is about 250 'C in the I atm case. Finally, the values

of fiber insertion plane to burning surface distance at N-,O appearance seen in the two figures are in

fact representative of the range seen in all cases. At 6 atm, interpolation in the point-by-point

record gives a distance of 0). 12 cm while using the linear fit through all position points Lives a

distance of'O. 13 cm. At I atm. the two distance values are 0.21 and 0.24 cm.

As mentioned above. Figure 4 is representative of the majority of the 6 atm cases. but

Figure 5 is not representative of most of the I atm cases, which instead look like the example in

Figure 6. Here, the N20 appears gradually. over a time period of at least a second, before

transmission is lost. The temperature and tlame-to-fiber distance ranges where transmission is lost

are not significantly different from those seen in the abrupt cases.

The I atm cases. with a much smaller change in transmission for the same N2O mole

Iraiction. usualllv had a signal-to-noise level like that shown in Figure 6. Without fitting i a smooth

curve to the ocbserved transmission it is difficult to see that there is a systematic decrease in

transmission corresponding to N2O appearance. The one I atm case in which N20) appearance is

easily visible, shown in Figure 7. is also one of two cases to show a second phenomenon: the

subsequent disappearance of N2O, again on the relatively slow time sclde of a few seconds.

This phenomenon was also seen in the 6 atm case shown in Figure 8. Once again, tme N20

both appears and disappears on a relatively slow time scale. Furthermore, a second. more abnipt

and higher concentration appearance of N-,O may occur just before transmission loss. (Note that

the data set of Figure 8 is clearly one of those taken before the addition of the N20) gas filter in

front of the reference detector. since both detector signals respond to the appearance of N2O.) The

cases in Figures 7 and 8 share two characteristics which may help to explain why only they show

to

1 500

E 0.9 V W ..... 450

S0.8 .- . ~gnai 400

-• 0.7 F• Last 35 )

0.57- 350a) a) 0.6-300 C-

LL N20"" 0 Appears , 250d 0.4 'V- :200 "•

E a-S 0150 E

M t nt - .-.9 ,7 0.2 ... ~1 0 •

rr = 0.1 -. --------- 50

1 61'.5 62 62.5 63 63.5 64 64.5 65

Time After Ignition, s

Ratio of Detector Intensities (N20/Reference). ........... Normalized N20 Detector Intensity

Normalized Reference Detector Intensity- --------- Embedded Thermocouple Reading, Degrees C

Point-by-Point Flame to Fiber Distance, cm-. - Linear Fit to Flame to Fiber Distance, cm

Figure 4. Example Data for XM39 Burning at 6 atm Pressure.

N2O disappearance: they both maintained good infrared transmission for several seconds after N2O

appearance. and their thernmcouple temperature readings in the steady state region were the highest

in each pressure set.

In other experiments, we found that the fluoride glass fibers can maintain high

transmission even when he~aed well above the softening point. The loss of transmission depends

on mechanical deformation, not on the fiber temperature per se. In the strand expefiments, another

factor would be misalignment of the two fibers once the surrounding material has melted. Thus.with improved support for the fiber it might he possible to obtain observations such as those in

Figure 7 and 8 routinely.

The final example. in Figure 9, was done to add support to the hypothesis that abruptly

appearing N20 is not evolved from the walls of the observation volume but arrives from regions

II

0.5 500

a)0.45 U 450

ýQ 0 .4 4.0....0,... .. " •7 '

- .0 -35 -• 350

Z 0.3- -- 300So 0.25 - " . . .. .. - .. ........

--- -:" ==.... --" 250 Co 0.2 - ----------- - --

0N .- o 0. 200E CL•' •0.1•'z !•' • • -"""" 15 E

z_•a.b 0.1 100

a: 0.05 50

0-130 131 132 133 134 135 136 137 0

Time After Ignition, Seconds

Figure 5. Example Data for XM39 Burning at I atm Pressure. (Same key as Figure 4). In thisrun, an N-2O gas filter cell was not used, so both detectors respond to N20.

0.4 ,,800

0.35 "• 700

. -0.3- 600

f)5 0.25 ..... 5000CU

LLi 0.2- k 400C30 D

500

N 0.15 300 Z

0.1 200 i-z F,•:.0ou.

0.05 .. 100

07 1 0126.5 127 127.5 128 128.5 129

Time After Ignition. Seconds

Figure 6. Example Observations for XM39 Burning at I atm Pressure. Showing Slower N-,OEvolution. (Same key as Figure 4.)

12

0.5 1000

0.45 .900

0.4 8'- .- . 800c, 0.5 7 00 v.

L 0.3- 00.25 ..... 0.0...

S 0.2- 400"i wJ

0.15 300 E

.Lo:' 0L . I --- 200

m 0.05 , 100

0o- C, .80 82 84 86 88 90 92 94

Time After ignition. Seconds

Figlure 7. XM39 Burning at I aim Pressure. Showing Slow Appearance and Disappearance of'N20O. (Same key as Figure 4.)

0.5 . , j500

m 0.35 350w---- ---- ---- ------- -----

S0.3- 300

0.25 250

~LLCL,...0.15. 150 F)

E 0.1 100

,.0.05 _ -------- - 250

0 3 . 4 .. .6 6

Time After Ignition. Seconds

Figure 8. XM39 Burning at 6 atm Pressure, Showing Slow Appearance nd Disappearance ofNE0. (Same key as Figure 4.)

o E t " :: '.

0.5 N-,. 04500

S? 0.4 •w• , • ,. r.,••.".400

V 0.35 350~-. v&~.. 40- 350

7ý 0.3 •,- 0o 0.25-"

- " 200 ",0.25 250al;

N~0.2-

a) 0.5-C

OE 0.15 N 150E.6 _ . 0.1 , 100 "

= a 0.05 ....... ---- 50

58 60 62 64 66 0

Time After Ignition. Seconds

Figure 9. XM39 Burning at 6 atm Pressure, With Kapton Film Inserted Above Fiber Gap.

closer to the flame front through suddenly opening cracks. In two cases, experiments were set upexactly like the series of 6 atm experiments which yielded sudden N2 0 appearance. except thatbefore the two segments of propellant strand were joined, a 0.0025 cm thick sheet of Kapton (astrong, high-temperature plastic) was placed on top of the fibers (covering about the central half of

the strand surface). With this more solid "roof', no abrupt appearance of N2,O was observed.Instead, the N20 effused slowly around the plastic sheet and into the observation volume.

4.2 Numerical Results

Table 2 lists averages or ranges of some of the parameters which can be derived from the

data traces exemplified by Figures 4 through 8. Beginning with the bum rates derived from thevideo records, it goes on to address five time points in the history of a bum: the appearance of N20

in the observation region, then two points at which the temperature history changes its functionalform, the point at which transmission goes to zero, and finally a point representative of conditions

near the flame front. The point-by-point flame front position histories were used to produce

distances between the flame front and the fiber at the time when the fiber plane temperature is goingthrough these points. Each of the points requires a few words of definition.

The definition of the N2O appearance point is straightforward for abrupt cases - for

slower filling cases we chose the point at which the ratio drops one standard deviation below itsmean for the previous 1M seconds. A criterion based on fitted curves could also be used, but those

14

Table 2. Averages and Ranges of Quantities Measured in XM39 Experiments

I atm (7 ca•ses) 6 aim (6 cases)

Burning Rate. cm/s 0.0150±0.00()14 0.0410±0.0025

N20O Appearance

N-,O Mole Fraction, Range (0.05-0. 19) (0.05-0.24)

Flame-to-Fiber Distance at N-,O Appearance Point. cmFrom Point-by-Point Measurements 0. 2 1+(). 1I ..16±0.12From Linear Fit to Flame Front Position ().26±0.11 (1.15±0.11

Thermocouple Readings at N20 Appearance, Range, 'C (240-410) (35-155)

Fill Time (from First N2 0 to Steady Value). Range, s (0.3-2.0) 15] ((0.1-0.3) [51(Numbers of Observations in Brackets) (0. 1-0.2) 121 (0.7) [I]

Steady State Time (for Short Fill Time Cases. Time (1-2) (0.3-0.7)During Which a Roughly Constant TransmissionIndicating N20 Presence is Maintained). Range, s

Atm, Flame-to-Fiber Distance/Burning Rate, s (8-28) (1.7-4.8)

Change in a Point in Temperature History

Distance to Flame Front at a Change Point, cm ().34±0. 17 ().()7±().()7

Temperature at ax Change Point. 'C 140±30 I l0-4()

Temperature History Inflection Point

Distance to Flame Front at Inliection Point. cm 0.26±0.085 ().()5±().()6

Temperature at Temperature History Inflection Point.°C 200±20 185±45

Slope at Temperature Inflection Point, 'C/s 35±15 400±20()

Transmission Loss Point

Flame-to-Fiber Distance at Transmission Loss. cm 0. 19±0.07 0.08±0.08

Tcmperature at Transmission Loss Point, 'C (250-500) (50-330)

Expected Time of Arrival of Flame Front at Fiber Plane

Distance to Flame Front at Att, after N-2O Appears, cm 0.06±0.)3 -00.2+0.)6

Temperatures at At,, after N2O Appearance, 'C 380±45 410±9()

Slope of Temperature History at Atff after N0., °C/s 10±6 1 00±4(0

15

points differed by less than I second from those derived as above. As will be discussed below.

the rise of temperature with time has an exponential form until, at an upper limit temperature. it

departs from that dependence. We define that point as the temperature history inflection point, and

defer a discussion of its significance to the following section, but for now we can note that even

with the often uneven burning in the fiber region. the average values of temperature and

temperature rate of change at this point have fairly small uncertainties. Furthermore, in the

exponential range of the temperature history, a discontinuous change in the parameters of the

exponential is observed. As will be discussed below, this is most easiliy ascribed to a physical

change in the propeliant which affects the thermal diffusivity (denoted by the symbol cc), so

Table 2 refers to the a change point, although a more neutral te-rm such as gradient change point

might be preferred. The transmission loss point is easily defined as the point where one detector

intensity came so close to zero that the ratio falls well outside of iLs previous range of variation.

The most straightforward way to derive a time for the last, flame front arrival event would

he to use the time at which the video flame front position history says the flame front is at the fiber

plane. However, we have already mentioned that these generally quite smooth variations of flame

front positions with time become more irregular near the fiber plane. It turned out that evaluating

average quantities at the zero flame-to-fiber distance time led to very wide ranges of variation. If

we assumed that the regression of the center of the strand was more regular, we could instead use

the linear fiLs to derive a time for flame front arrival. In Table 2 we use a third method,

extrapolating from a point where we expect the flame front and burning rate to be relatively

unperturbed. This point could be defined in a variety of ways, but we chose it to be the N2O

appearance point, simply because its flame-to-fiber distance was of intrinsic interest and already

had been tabulated. This distance and the burning rate are used to derive a time between N2O

appearance and the arrival of the flame front. At.f, reported in the N20 appearance section. This

time is then added to the N20 appearance time to generate a time at which to evaluate properties

which may represent conditions near the flame front. Simply as a check, we evaluate the position

relative to the flame front from the video record for these times. As can he seen in Table 2, on the

average the video flame front is very close to the fiber plane at AtT from N2 0 appearance for the 6

atm cases, but appears to be slightly above the fiber plane in the I atm cases.

It is of interest to examine the flame-to-fiber distance entries in Table 2 to establish the

sequence of evenLs experienced at the fiber insertion plane. The sequence is different at the two

pressures. At I atm. the temperature inflection point can arrive first, followed in fairly rapid

succesion by the appearance of N2O and the loss of transmission, with the arrival of the flame

front occurring considerably later. At 6 atm, the N2 0 appears first, substantially before the arrival

1 ih1 101np10liattLrc ifl ieCtion point. and rhomrlorc in a lowcr tInipcratUrC rcgimc11. I . .,S I

Irian' Issi onl. inlelction Point amval, and all-ival ot the flame [roni region a& dc li ned by the above

procedure. all occur at about the same time. In the lolowving sectIon. we will sh nythat these

cliatutes in the e,,quence of events tollow naturally Irom the dependence of the thermal wave on the

burning rate and the assumption that the length that NA() can travel through the propellant is less

dependen on the birning rate and more i constantI of thie material.

17

5.(0 THERMAL WAVE ANALYSIS

Before we can hypothesize about the source of the N2O we observe, we must understand

why it appears in two different temperature ranges for the two pressure regimes. We can analyze

the thermocouple temperature records using a model which utilizes information on the

thermophysical properties of the propellant. This is of interest not because it provides accurate

measurements of these properties (again, individual bums showed wide variations in almost all

observed parameters), but because it will be found that the thermophysical properties derived from

the average parameters lor all bums are roughly consistent with literature values, and so the

expected time/temperature history based on those known thermophysical parameters can be used to

interpret our results.

It has long been known13 that solution of the heat balance equation for a thermal wave with

it constant surface temperature progressing into a solid having constant thermal properties and no

heat sources or sinks leads to an expression for the steady state temperature profile which has the

generalized form.

(T-To) = (Tj11t1-To)exp(-(x-Xo)/6). (I)

In this expression. x is the (positive) distance into the propellant from the flame front. T is the

temperature at x. and To is the ambient temperature. From the derivation. Tinllt is the temperature at

which the thermal wave departs from exponential behavior, which occurs at the plane where heat

production ( ) becomes important. The effect of this heat source is seen in the shallower

temperature gradient in the non-exponential region. a gradient extending further into the solid due

to heat production. In the early work on double-base propellants. Tj1j could be identified with the

burning surface temperature T, It will be seen that this is not the case for the composite propellant

studied here. since Table 2 shows that a departure from exponential behavior occurs at a much

lower temperature than the surface temperature. This difference results in a second modification to

the exponential expression. the presence of the parameter X0 which is the distance between the

flame front and the temperature history inflection point. Finally, the length scale parameter 6 is

related to the thermal diffusivity cc and the burning rate rh by

6= Wtrb (2)

Equation 2 represents a competition between the ability of heat to move through the solid and the

speed of the flame consuming the solid. If the former is large, 8 is large and heat penetrates a long

18

distance into the solid. When the hurning rate is relatively large. 6 is small and the temperature

gradient extends only a short distance into tile propellant. The thermal dilfusivity is in turn related

to the thermal conductivity X, and the heat capacity cp by oX = •/Pcp where p is the density. Our

observations. of course, do not shed any light on these quantities individually.

The applicability of a simple model becomes clearer when the temperature with time data

are displayed in a semilog plot. as in Figures 1() and 1I. It is seen that simple exponential forms lit

very well over two temperature regions. The parameter 6 is derived from such plots by dividing

the average burn rate by the slope of each temperature versus time curve. Therefore. to begin by

representing the temperature histories in the smallest number of parameters. we fit two exponential

length scale parameters. 6. to each case, then cemputed a mean and a standard deviation for each

pressure and temperature regime. These parameters are reported in Table 3. Then we used the

inflection point temperatures and positions reported in Table 2 to solve for a pre-exponential factor

C'. in

(T - To) = Cexp(-x/6) (3)

These parameters are also reported in Table 3.

A point which should he mentioned now is that the assumption that the change in gradient

parameters between tile two temperature regions corresponds to a discontinutous change in the

thermal diffusivity (perhaps due Simply to bubbles or cracks appearing at the transistion

temperature) is only one possible explanation. Another is that the steep exponential region between

the two transition temperatures is due to an endothermic process such as melting or boiling. The

dlrawback to this alternative is that the exponential form would not be expected to hold. without a

fortuitous dependence of the rate of heat loss on temperature.

If we keep the hypothesis that the transition between two exponential gradients involves a

chang'e in thermal dilfusivitv. we must next address the fact known from other work that the

thermal difilusivitv is a IUnction of temperature. Experiments in tile -20 to 50) "C range have yielded

at three parameter lit to W(I) 14 . Having shown that we can represent our observed data in terms of

simple exponential forms, we now want to show that those data are consistent with a plausible

temperatture dependent thermal diffusivity. For a temperature dependent ox. the closed form

Equation I does not obtain. However, it will be sufficient to approximate the true (non-closed-

iorm) solution by Ftuation I with a temperature dependent (x, and to use a linearly temperature

dependent oxWT). ot(T)=(x0 + (L1T, because all we want to do is show that our data are

19

102

"JJ 10'

a

10-158 59 60 61 62 63 64 65 66 67 68

Time after Ignition, Seconds

Figure 10. Temperature History of 6 atm Case Shown in Figure 4, Showing Fits to SimpleExponential Expressions (Dotted Lines).

102-

10i

1010 115 120 125 130 135 ,140

Time after Ignition. Seconds

Figure I11. Temperature History of I atm Case Shown in Figure 5, Showing Fits to• SimpleExponential Expressions (Dotted Lines).

20I

Table 3. Parameters Obtained from Thermocou~ple Ternperature Histories

Exponential Fit to Low Temperature Range 1 aitri6at

6. Length Scale Parameter, cm 0.165±0.032 ().042±0).0)(65

C. Pre-exponcntial Factor (Uncertainty Ran-ge).0 C 595 (430-620) 330 (185-430)

Exponential Fit to High Temperature Range

6. Lena~h Scale Parameter, cm 0.095±0.044 ().()24±0).00(9

C. Pre-exponential Factor (Uncertainty Range), 'C 2680 (1260-25200) 1090 (825-2470)

Model Parameters Consistent with Exponential Fit

Tillfl-T 0. 2C 174 160

Thermal Wave Origin. X0, cm 0.26 0.045

Lo~k TemperatUre Range2

Thermal Dit'fuLSiVity Constant Term. uo, cm 2/s 0.001I9

u Temnperalture Coefficient, u.l. cm 2/s 0C-1 9 x I o-()

High Temperature Range

Thermal Ditfusivity Constant Term, u.(), cm 2/s 0.0014

(x Temperature Coefficient, (xj cm 2/s C-1 2 1 I -6

21

noisy enough that it is difficult to distinguish between a constant and a temperature-dependent

thermal diffusivity.

To do this, we use the Equation 3 representations of average temperature histories, together

with upper and lower error limit expressions derived as follows. The 6 parameter in the upper

limit expression is the average 6 plus its standard deviation. The C parameter for the high

temperature range is then obtained by fitting to the inflection point temperature plus its standard

deviation. The C parameter for the low temperature range is set by matching the high temperature

expression at the (x change point. Since the 6 parameters, which appear in the exponential. have

fairly large standard deviations, and since the range of observed temperatures at any distance from

the flame front is relatively small, the C parameters are forced into some large variations, as seen in

the ranges reported in Table 3.

To show how these exponential forms compare with some individual observations,

Figures 12 and 13 once again exhibit the temperature data for the two cases shown in Figures 1()

and I I. Here. however, the video flame front position record has been used to convert the time

axis into a distance from flame front axis. The exponential data representations using the

parameters in Table 3, and the uncertainty limit expressions derived as discussed above, are also

plotted in the figures. It can be seen that the substantial variations in the distance scales cause these

observed curves to fall outside the error limit expressions. In the 6 atm case, the observed

inflection and x change points occur at (). 10 and 0. 13 cm from the flame front, much further out

than the average values of 0.05 and 0.07 cm. In the I atm case. on the other hand, inflection and

ct change point at 0.22 and 0.26 cm are well inside the average values of 0.26 and 0.34 cm.

However, the uncertainty in the position axis does not directly affect the determination of thermal

diffusivity values from the slope of the temperature history curve.

Now that we have a compact way of expressing what, on average, the temperature histories

were in the two pressure regimes, we can go on to compare to expected temperature histories as

predicted by Equation I. In Figures 14 and 15 we present the Equation 3 exponential data

representations as dashed lines, with the error limits expressions again yielding two bracketing

dotted lines. The solid lines are given by Equation I using the parameters in the lower hall of

Table 3. The pre-exponential factor C can be divided into the inflection point temperatures given in

Table 2 and exponentials involving X0 parameters which are essentially the average observed

positions of those inliection points. The linear or(T) expressions given by the parameters in

Table 3 are those which minimize the squared differences between the dashed and solid lines in

Figures 14 and 15. These linear thermal diffusivity expressions are plotted in Figure 16, along

22

102

o102

100

101

10".3 0.25 0.2 0.15 0.1 0.05 0 -0.05

Distance from Flame Front, cm

Figure 12. Thermocouple Reading versus Flame Front Position of 6 atrn Case Shown in Figure 4(Solid Line), Average Exponential Data Representations (Dashed Lines), and ErrorLimit Expressions (Dotted Lines).

103

a)0~

-• 0 •............. oo ....... ...""

E

10.10.5 0.45 0.4 0.35 0.3 0.25 02 0.15

Distance from Flame Front. cm

Figure 13. Thermocouple Reading versus Flame Front Position of I atm Case Shown in Figure 5.(Line types have same definitions as in Figure si3.)

23

102 ..........

.........

...........

.......... .o .o .o O . O -..

.. y...y . y.....' " " ".."" .

10 1 ......... .. ,,'

CD

1000.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04

Distance from Flame Front, cm

Figure 14. Comparing Simple Exponential Data Representation and Error Limit Expressions for6 atm Cases (Dashed and Dotted Lines) with Equation I Expression IncludingTemperature-Dependent Thermal Diffusivity and Other Table 3 Parameters (SolidLines).

103

U

--\1 1 0 2 -..... ".........

S . . . . . . , -' " . . . . . . . . . . . . . , , " " " " . . .. . .. .... .... .... . ...... ..

2•,d .............

10.5 0.45 0.4 0.35 0.3 0.25

Distance from Flame Front. cm

Figure 15. Comparing Simple Exponential Data Representation and Error Limit Expressions for

I arm Cases (Dashed and Dotted Lines) with Equation I Expression Including

Ternperature- Dependent Thermal Diffusivity and Other Table 3 Parameters (Solid

Lines).

24

C"'

S1.5 .

----------------------- -------------------------->U,

"Ea)

S0.5

20, 40 60 ; 100 .20 140 160 180 20C

Temoerature. Degrees C

Figure 16. Thermal Diffusivity for XM39: Low Temperature Measurements1 4 (Solid Line), DataPoint from Strand Burning Observations of Reference 15 (With Quoted Error Limits,Plotted at Observed Inflection Point Temperature of 165 'C), and Our Linear or(T)Expressions From Table 3, Used in Figures 14 and 15. (Dot-Dash and Dashed Lines).

parameters which give the best match between Equation I and Equation 3 expressions while still

being roughly compatible with the other measurements of (x. many other or(T) values would also

fit our observations well enough. This is shown in Figure 17, which shows a region of oX(T)

values which give temperature history curves which lie within the error limit expressions for both

the I atm and 6 atm cases. In other words, the uncertainty in what values of ox(T) are consistent

with our observations is at least as large as indicated in Figure 17. Considering these large

uncertainties, it is not a matter of great concern that our average ax values at low temperatures seem

to be higher than those of Reference 9. nor that our high temperature ox values seem to be below

that reported in Reference 15.

To summarize, a single set of W(T) expressions can be used to reproduce the average

temperature histories for both the I atm and 6 atm data sets. This is in spite of the fact that (at the

simple exponential level of data analysis) the average 6 parameters for the two data sets are not in

the ratio of the burning rates for the two pressures (rh( 6 atm)/rh( I atm) = 2.7 ± 0.3), but instead,

the ratios in both temperature regimes are about 4. However, the uncertainties in the ratios of

6(6 atm) to 60( atm) are large (±+.() and 2.4 for low and high temperature ranges), s,, much of the

difference can be put down to large standard dcviations, with a component also potentially

ascribable to the temperature dependence of ax and the fact that somewhat different ranges of

temperature contribute to single values of ax derived from the two different pressure cases.

25

To summarize, a single set of Ox(T) expressions can be used to reproduce the average

temperature histories for both the I arm and 6 atm data sets. This is in spite of the fact that (at the

sim ple exponental level of data analysis) the average 6 parameters for the two data sets are not inthe ratio of the burning rates for the two pressures (rb( 6 atm)/rb( I atm) = 2.7 ± 0.3). but instead.

the ratios in both temperature regimes are about 4. However, the uncertainties in the ratios of

6(6 atm) to 6(1 atm) are large (±1.0 and 2.4 for low and high temperature ranges). so much of the

difference can be put down to large standard deviations, with a component also potentially

ascribable to the temperature dependence of (x and the fact that somewhat different ranges of

temperature contribute to single values of c( derived from the two different pressure cases.

)(0-32.5

n 2-

U

('I.-, 1 5. . .

° ....,,

E

S0.5

0 20 40 60 33 100 120 140 160 "80 20C

Temlerature. Degrees C

Figure 17. Thermal Diffusivity for XM39: Low Temperature Measurements 14 (Solid Line), DataPoint from Strand Burning Observationsl-, and Range of Values whose Equation ITemperature Histories Fall within the Error Limit Regions in Figures 14 and 15 (Dots).

26

(.) DISCUSSION

This study has produced three major results. The first is that we have observed N-,O. a

gaSCO us decomposition product of RDX. evolving into a small volume in the condensed phase of aburning nitramine propellant strand. As we will discuss below, this result is expected. in that it isconsistent with other observations of the chemistry of RDX decomposition. The second is that in

the I atm cases, there is strong evidence that N20 is not seen until substantially after the RDX

melting and decomposition processes have begun. It has been expected that N2O is not a primaryproduct of RDX decomposition. nut it is somewhat surprising to see such a clear separation. The

third result is that many 6 atm cases show an abrupt N20 evolution at a time when the propellantforming the observation volume is still at a temperature well below that needed to produce rapid

decomposition of RDX. We can suggest a mechanism for this observation, but first we need toplace it in the context of what we and others have learned about XM39 burning and RDX

decomposition.

6. 1 Correlation of Observed Events with Temperature History

To aid in visualizing the relationships between the events listed in Table 2, in Figure 18 wepresent a sketch of temperature histories and event positions and temperatures which are roughly

consistent with those in the table. It turns out that it is not possible to construct a self-consistent

graph out of the averages listed in Table 2, because the large variances and small numbers of pointsmean that different runs have different weights in determining the various parameters. Even suchbasic qualitative characteristics as the order of events cannot be completely summarized by a single

I Igture. For example, although four 6 atm runs do indeed have the order shown in the figure,N20) appearance/transmission loss/ox change/inflection point, two others have a different order.

c, change/N20 appearance/inflection point/transmission loss. Still, to the extent that there is atypical set of positions and temperatures associated with the set of ohserved events, Figure I 8 is as

good a representation of it as our present data set will allow.

In the context of Figure I 8, then, we want to ask what we know about the state of thepropellant through the time interval surrounding the observation of N2O appearance in the

observation region, based both on our data and on other sources. We begin by noting that in

Table 2. three of the events are associated with temperatures with fairly small standard deviations,

suggesting that they involve processes which are legitimately associated with specific temperatures.

The burning surface. of course, is expected to have a temperature which is only secondarily a

27

TRANSMISSION LOSSýd N20 APPEARANCE,&

INFLECTION POINT

ILI

102 CHANGE I I

TRANSMISSION LOSSIL

ILI 1 ATMOSPHERE

N20 APPEARANCE6 ATMOSPHERES

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

DISTANCE FROM FLAME FRONT, cm

Figure 18. Qualitative Sketch of Typical Thermal Waves and Positions of Other Events in

Strand Burning Observations

function of combustion pressure (although our strand burning in air is certainly expected to have ahigher surface temperature than observed in inert atmospheres). The value from our extrapolation

procedure for the time of arrival of the flame front. of about 400 °C, is consistent with the rangequoted by Brill and Brush16 of 350-400 °C for the surface temperature of burning HMX and RDX.

Again, the temperature inflection point, occurring around 190 °C, is thus not to be

identified with the burning surface temperature. This conclusion is reinforced by the fact that in Iatm cases we have good infrared transmission over a long path through a channel in the solid long

after the inflection point has passed the observation region. We know the maJor component, RDX,

does not decompose at significant rates until it has melted. Consulting Table 1, we find that RDXmelL's at 190 'C. (This value is a manufacturer's specification--the references to values in the 203-205'C range may refer to material of higher purity than would be found in gun propellant.2 ,•'"7 .)

Therefore, our inflection point may be associated with the melting, decomposition, and eventual

heat releasing processes in RDX. It is clear that if RDX melting begins at the inflection point

temperature, its effect on the temperature history is masked by the exothermic process. An

29

altern~aLive explanaton. that what we have referred to as the (x change point is actually the RDX

melting point, would shift all our observed temperatures to surprisingly high values.

Because these processes take place at finite rates, as the burning rate increases we expect

the inflection point to shift to higher temperatures. as seen in the observations of Reference 15.

These covered the pressure range from 75 to 75(0 psig, and showed a strong dependence of the

inflection point on pressure The temperatures reported at the two extreme pressures are about 165

and 370 'C, respectively, while at 2(X) and 5(X) psig the temperatures are 250 and 300 'C. It can beseen that the expected variation from I to 6 anm is smaller than our standard deviations, so our

failure to observe this trend is not surprising.

Finally, we want to ask if the lower temperature change in the form of the thermal wave.which we have referred to as the ox change point, can he identified with a property of one of thepropellant ingredients. If we note the temperature ranges in Table 2. centered around 125 °C. and

then consult Table I, we find the closest temperature listed to be the boiling point of the plasticizer.Acetvl triethyl citrate is a liquid at room temperature. and its plasticizing action essentially involves

partial solvation of the other propellant ingredients. The thermal properties of this solution will be

different than those of the pure liquid plasticizer, but it is not unreasonable to expect a sharp changein properties near the plasticizer boiling point, due to either vaporization or themial decomposition.Here again it must be remarked that in their observations of strand burning of XM39 with

embedded thermocouples, neither Reference 9 or 15 noted such a point in the temperature history.We can only suggest that at their higher pressures and faster time scales, the process involved is

too slow to be observed as a sharp discontinuity. This point too is not seen by us to move towards

higher temperatures at faster burning rates. but again our error limits are substantial. We should

also note again that the steeper gradient between inflection and ot change points may not be a

thermal diffusivity effect at all. but simply reflect a heat consumption process such as melting,

boiling, or thermal decomposition.

Having made the case that three of the events identified in Table 2 and Figure I8 are, under

our conditions, correlated with relatively fixed temperatures, what can be said of the other two

events? It is clear that transmission loss is not correlated with temperature, and thu wide variationin temperatures at which it occurs tells us that a number of different processes can be involved in

bringing the transmission to zero, including misalignment of the fibers and intrusion of moltenpropellant material into the observation region. It turns out, however, to be fairly well correlatedwith distance from the flame front, with the standard deviation in position for the I atm cases being

particularly small.

29

Finally. it is clear from IFigure IS that appearance of N-,() in the observation voluime i,,, not

correlated with the temperature at that point. If anything, it too is better correlated with distance

from the flame front. To understand why we suggest that the N-,O we observe has been lorned

not in the observation region, but substantiallv nearer the flame front, we need to review what we

know about combustion of* nitramine propellants like XM39.

First, we know that N2O is a product of thermal decompposition of RDX. but not a final

product of the gas phase combustion of those solid decomposition produtcts. either in tile primary

flame or in secondary combiustion with added oxygen. Tihe evolution oft N)0 from heated RDXhas been observed by a number of investigators. using mass spectrometric and infrared absorption

detection techniques. 3,6.' I,1.I18 An example of the quantitative measuremenLs of Fetherolf and

Litzincer1 I is given in Figure 19, along with our estimate of how the somewhat broader range of

species observed by BehrenstX in his low pressure experiments would evolve under the I atm

conditions of Reference I I. The point of this figure is that under a wide range of conditions N-,()

is expected to be a major gas phase product of the thcrrnal decomposition of RDX. The second

point, that N2O is not expected to be a major product of gas phase combustion of RDX

decomposition products. is made by Figure 20. It shows results of the gas phase chemistry

calculations of Melius 19.2o for RDX burning in I atmosphere of an inert -as. In his mechanism.

N20) is consumed by thermal decomposition as well as reaction with hydrogen atoms, and has

essentially disappeared 0. 1 cm above the burning surface. Thus. we know that the gas we are

observing comes from inside the condensed phase of the propellant, not from the exhaust gas from

the flanme. In any case. the rapid gas flow over the strand prevents recirculation of the exhatist cas

into the fiber insertion region.

We also know that measurements of the kinetics of RDX decomposition show that gas

eVOlutition is very slow below thie melting point. 5.1 .21 Yet we also know that infrared transmission

through even 1). I cm of molten propellant will be essentially zero, so that the walls ol the

observaion region must be, at least, below the melting point of the CAB binder (215-225 "

Further. we know that the N20 appearance rate in cases like those of Figures 4 and 5 is faster, by

one to two otrders of magnitude, than could be expected from N2() evoluition from the walls of the

observation region. when the observed evolution rate from Reference IS for RDX at 205 `C is

used. All these facts lead us to suggest that N,0 from RDX decomposition accumulates In

pressurized bubbles in regions of the propellant strand which are hotter and closer to tile flame

front than the observation region. and that the N2() is transported from the pressurized buibbles to

the observation volume by the opening ol cracks connecting the two regions.

10O

" 2-

o 1

U-

.15

C 05 ..........-

N,0 H2CO NO2 HCN H20 NO CO, H2

Figure 19. Major RDX Decomposition Gas Species from Experiments ot Fetherolf andLitzinger.1I Compared to Adaptation of Observationsof Behrens (dots)18

)4

--.35-

"23

o .2 25

L- 0 2

0

D.15

01 -

CO H,0 NO CO2 H,)005

N2O NO2 HCN

Figure 20. Major Gas Phase Products of RDX Primary Combustion, from Model Calculationso1' Melius. 19. 20

3'

Sveral roulps have ll Made .canning electroll microscope (S EN,) ohservations oI sections

through the surface of quenched strands of X.M39. 15 A ni.\elt region is observed. ol thickness

S0)1 -0).(03 cm or burni ng in air at 1 atmosphere pressure. This melt region is filled with hubbIes

Ot varIous swes. rarwinm in diamnetcr up to th1e lull thickness of the melt laver. Because these

Nubbles are so close to tile burning surface. it would he surprising if they were the source of N•O

observed on the order of 0. 1 cm further down, in relatively still-solid propellant. WO would expect

most of these bubbles to burst upward. ejecting decomposition gas into the I lame. and indeed what

we could see of our huming surfaces seemed to he consistent with a melt laver containing growing

and bursting bubbles. Furthermore. a single bubble, even with a 0.03 cm (3(00 pm I diameter, will

not fill the observation region. a cylinder of 0.02 cm diameter and order of ().3 cm length. to the

mole fraction levels we observe, which are consistent with pure decomposition gas at ambient (I or

0 atrn pressure.

These objections lead to the suggestions that tile bubbles, or pressurized voids, which are

tile source of the N-,() we observe. are not in the melt region but below it. in a recion where at least

the higher-melting binder has kept enough strength to hold these bubbles in place, and that our

observations oft sudden filling of the observation volume with N-,O involves tile nearly

simultaneous opening of several cracks connecting it to several bubbles. Such cracks have not. to

our knowledge. been observed in any SEM photographs of XM39 strands. but there need not be

many of them. and they will be more difficult to see than the bubbles. Finally. we cannot at this

poi1nt rule out tile possibility than our fiber insertion procedure. involving pressing a groove at least

(M)1 cm deep into tile propellant, has generated fractures extending many times that depth into the

",lid. or has at least set up strains which make crack formation mtore likely during burning.

Co rrelation of Our N,() Observations with Chemical Mechanisms of R[X Decomposition

This last part of our discussion has to do with the status of N2() as a major decomposition

product o1 RDX. It tunis out that tile observation of substantial NO() is completely consistent with

the existing literature of experiments, theoretical predictions. and analysis. To show this. we

present a brief review of some of the major contributions to tile current overall picture of RDX

dc.omp•ostp)ion.

It has been widely noted 16IX. 20 that there appear to he two global decom position channelsfor RI)X. one leading to the products N2() + il')O, and one leading to NoV + I ICN (or

I ION() + I ICN. or No)2 + I l('N + [H. the differences expressing the fact that some of' the products

f this, channel are less stable). The pathways to these products can include a number of parallel

.12

processes. some of which can involve a number of sequential steps. Not surprisingly. then. some

variation in relative amounts of the above products and others has been ohserved depending on the

conditions durnng the decomposition. Experiments on RDX decomposition Linder relatively

controlled conditions. including the use of deuterium isotope techniques, have unraveled some of

the details of these processes.

Behrens and BulusuIX have been able to explain their simultaneous themogravimetry

modulated beam mass spectrometry observations of RDX thermal decomposition under high

vacuum conditions using four pathways:

I. Formation of I -nitroso-3.5-dinitrohexahvdro-s-triazine (ONDNTA - RDX withone NO 2 group replaced by NO) followed by ring breaking to yield N)O. CH20.and other products.

2. HONO elimination to form oxv-s-triazine (OST). a molecule which retains the C-Nring of RDX and has the formula C3 H3N30 and an uncertain structure, as well asH-2O. NO. and NO-,.

3. N-NO-, bond breaking leading to ring breaking and formation of NO-, + H-,CN +2N-,O-+ 2CH 2,O.

4. Catalytic decomposition to fomi N2 0, CH 2O, NO 2 and NHCHO.

They find no evidence that process 4 is the water catalysis suggested by Melius 2 . and instead

suggest the catalyst is a residue of RDX decomposition which builds up on the walls of their

reaction cell. That makes this mechanism specific to their apparatus. although analogotIs processes

might Occutr in the burning of an RDX composite propellant. The other three mechanisms all

involve N-N bond breaking, since process I is thought to proceed by N-NO1 bond cleavage

followed by reaction with a source of NO.'9 This might seem to be in conflict with the deuterium

kinetic isotope effect studies of Shackelford et a1.30.31 who found C-H bond rupture to he the rate-

controlling step. However. Melius 28 has pointed out that not only does N-NO 2 bond breaking

weaken remaining bonds. but C-H bond breaking results in a negative N-NO2 bond energy with

only a 2 kcal-mole-1 harrier for leaving. Furthermore. Behrens and Bulusu found evidence of C-H

bond breaking in both pathways I and 2. which together result in two-thirds of the total RDX

decomposition in their experiment. Behrens and Bulusu also note that pathway 3, whose

contribution is estimated to be only 1()'/, of the total. may be the concerted symmetric triple fission

to form three methylenenitramine fragments observed in the gas phase by Zhao (,t I. 32, or may

instead be a process yielding two methvlenenitramine molecules plus H-CN and NO-.

33

To summarize.[ the above experimental evidence seems to give a coherent picture which

agrees with the prediction of Melius2 8 oi energetic grounds that RDX decomposition in the

condensed phase will he dominated by processes involving N-NOn bond breaking, at least at high

heating rates where catalytic processes are not important. The dependence of processes and

product distributions on heating rates is part of a more general dependence on the environment of

the RDX molecule. including not only the temperatures and time scales involved but also the phase

(process I is thought to be dominant in the solid phase. and at low temperatures in general).

Table 4 gives a review of experimental observations of RDX decomposition to give Ni0 ordered

by heating rate. It can be seen that the only experiments in which conditions did not allow

progress beyond the initial NO 2 loss step are those of Wight and Botcher2 5--7 (although with

longer laser pulses they could observe formation of other products including N20). As to the rest

of the experiments, covering five orders of magnitude in heating rate. it is not easy to discern any

svystematic trends in product distributions. given that differences in observed concentrations are

only at the factor of two level and that other differences in the experimental conditions can also

ea.sily generate differences at the same level. Our experiments (with heating rates taken from the

temperature history slopes in Table 2) are seen to he in the middle of this range of heating rates.

We can add the comment that the nature of the initiating process for RDX thermal

decomposition has been particularly controversial because of a seeming disagreement between the

above-mentioned results of infrared multiphoton dissociation experiments by Zhao et al. 32

showing that two-thirds of their RDX decomposed by a concerted detrimerization process. and the

condensed phase experiments which find little or no evidence for such a process. However, the

sugeestion has been made by Wight and Botcher 27 that in fact no discrepancy exists, that while the

concerted process proceeds easily in the vapor phase. the condensed phase constrains the motion

of the RDX molecule too much for it to contribute significantly. Thus the picture of RDX

decomposition in the condensed phase remains one in which all ring-breaking proceeds by several

sequential steps and results from the destabilizing effects of NO2 loss.

Our experimental observations of N2O are consistent with the above picture. Our heating

rates are slow compared to chemical reaction rates, and our N0 originates in a mixed gas/liquid

region with plenty of opportunity for homogeneous or heterogeneous reactions to carry initial

products of decomposition to secondary products. A decomposition product whose observation

would tell us more about the environment in the decomposition region and during their transport

through the propellant is NO-, which is expected to be more reactive as well as being formed by

sonme of the first bond-breaking processes in RDX. Therefore, in ongoing work, we plan to look

.34

lor NO, simultaneously with N2O. with the other experimental parameters remaining much the

same.

Table 4. N2O Formation at Various Heating Rates

Research Group Technique Ref. Heating Rate. K/s N2 Q0 bservC•d

Karpowicz and Brill Rapid-Scan FTIR Spectroscopy 2 2.5 Yes

Oyumi and Brill In Situ Rapid-Scan FTIR 3 8-200 Yes

Brill et al. SMATCH/FTIR 23 10M-350 Yes

Brill and Brush T-Jump FTIR 16 < 2000 Yes

Behrens and Bulusu Thermogravimetry/MBMS 1 0.017 Yes

Pesce-RodrigueilFifer Pyrolysis/FTIR 24 0. 17 Yes

Womihoudt et al. Strand Burning/IR Absorption This I()-1_() Yes

Work

Wight and Botcher Laser Heating/Cryotrap/FTIR 25 1-3 x 107 No

35

7.(T SUMMARY

We draw three major conclusions from this work: N2O is observed as a decomposition

product nf RDX under actual propellant combustion conditions. it is formed in high temperature

regions of the condensed phase and is able to penetrate into regions of the propellant which areotherwise little affected by the advancing flame front, and under our conditions its fomiation is notan initial step in RDX decomposition. Of course our observations by themselves will not fully

specify the structure of a burning strand and the physical mechanisms which lie behind it, but they

certainly strongly encourage the speculation that the suddenly appearing N20 originates inpressurized bubbles inside liquid RDX which in turn is contained inside a matrix formed by the

propellant binder. This matrix would have to be structurally sound enough that some bubbles areprevented from releasing their contents upward into the flame region. but instead eject

decomposition gas into the observation volume when cracks open creating vent paths.

36

8.0 REFERENCES

1. M.H. Alexander. P.J. Dagdigian. M.E. Jacox. C.E. Kolb. C.F. Melius. H. Rabitz. M.D.Smooke. and W. Tsang, Prog. Energy Combust. Sci. 17, 263 (1991).

2. G.F. Adams and R.W. Shaw. Jr.. Ann. Rev. Phys. Chem. 43, 311 (1092).

3. R.J. Karpowicz and T.B. Brill. Comb. Flame 56, 317 (1984).

4. Y. Oyumi and T.B. Brill, Comb. Flame 6.2, 213 (1985).

5. P.J. Miller, G.J. Piermarini, and S. Block. Appl. Spectrosc. 38, 680 (1984).

6. P.J. Miller. S. Block and G.J. Piermarini. Comb. Flame 8-3, 174 (1991).

7. J.P. Toscano and M. McBride. in ONR/LANL Workshop on the Fundamental Physics andChemistry of Combustion, Initiation, and Detonation of Enegetic Materials, CPIAPublication 589, The Johns Hopkins University, March 1992. p. 7 1.

8. T.B. Brill, Prog. Energy and Comb. Sci. .8, 91 (1992).

9. M. Miller. 26th JANNAF Combustion Proceedings, CPIA Publication No. 529, Vol. III.1989. p. 343.

I10. K.L. Erickson, W. M. Trott, and A.M. Renlund, "Use of Thin-Film Samples to StudyThermal Decomposition of Explosives", paper presented at 18th International PyrotechnicsSeminar, Breckenridge, CO. 13-17 July, 1992.

I i. B.L. Fetherolf and T.A. Litzinger, The Pennsylvania State University. privatecommunication, 1992.

12. L.S. Rothman, R.R. Gamache. R.H. Tipping. C.P. Rinsland, M.A.H. Smith. D.C.Benner.V. Malathy Devi, J.-M. Flaud. C. Camy-Peyret, A. Perrin. A. Goldman. S.T.Massie. L. R. Brown and R. A. Toth. J. Quant. Spectrosc. Radiat. Transfer 48, 469 (1992).

13. R. Klein. M. Mentser. G. von Elbe. and B. Lewis, J. Phys. Colloid. Chem. 54. 877(1950).

14. M. Miller, U. S. Army Research Laboratory, Aberdeen Proving Ground. privatecommunication. 1993.

15. W.H. Hsieh, W. Y. Li and Y. J. Yim. AIAA Paper 92-3628. presented atAIAAISAE/ASME/ASEE 28th Joint Propulsion Conference, July 6-8. 1992

16. T.B. Brill and P.J. Brush, Phil. Trans. R. Soc. Lond. A 339, 377 (1992).

17. Melius. C. F., 25th JANNAF Combustion Proceedings, CPIA Publication 498, Vol. II.1988. p.155.

18. R. Behrens, Jr. and S. Bulusu. J. Phys. Chem. 96, 8877 (1992).

19. C.F. Melius, 25th JANNAF Combustion Proceedings, CPIA Publication 498. Vol. 11.1988, p. 155.

37

20. ('F. Melius. in Chemistry and Physics of Energetic Materials, Ed. S. Bulusu. ASI 309,

NATO, p. 51.

21. L. Huwei and F. Ruonong, Thermochimica Acta 138, 167 (1989).

22. M.A. Schroeder, R. A. Fifer. M. S. Miller and R. A. Pesce-Rodriguez. BRL-MR-3845.Ballistic Research Laboratory, Aberdeen Proving Ground. MD. June 1990.

23. T.B. Brill, in Chemistry and Physics of Energetic Materials, Ed. S. Bulusu. ASI 309,NATO. p. 255.

24. R.A. Pesce-Rodriguez and R.A. Fifer. Pure & Appl. Chem. _6_, 317 (1993).

25. C.A. Wight and T. R. Botcher, J. Am. Chem. Soc. 1_.4, 8303 (1992).

26. T.R. Botcher and C.A. Wight, in Structure and Properties of Energetic Materials, Ed. D. H.Liebenberg, R. W. Armstrong, and J. J. Gilman, Mat. Res. Soc. Symp. Proc. 296, 47(1993).

27. T.R. Botcher and C.A. Wight. J. Phys. Chem. 97, 9149 (1993).

28. C.F. Melius, in Chemistry and Physics of Energetic Materials, Ed. S. Bulusu, ASI 309,NATO. p. 21.

29. R. Behrens, Jr. and S. Bulusu, J. Phys. Chem. 96, 8891 (1992).

30. S.A. Shackelford. S.L. Rodgers, and R.E. Askins. Propellants. Explosives andPyrotechnics 16, 279 (1991).

31. S.A. Shackelford, in Chemistry and Physics of Energetic Materials, Ed. S. Bulusu.ASI 309, NATO, p. 413.

32. X. Zhao, E.J. Hintsa and Y.T. Lee, J. Chem. Phys. .. , 801 (1988).

38

Q.0A LIST OF PUBLICATIONS AND TECHNICAL REPORTS

"Infrared Fiber Optic Diagnostic for Solid Propellant Combustion," J. Wormhoudt and P.L.Kebabian. in Structure and properties of Energetic Materials, Materials Research SocietySymposium Proceedings Vol. 296, Eds. D.H. Liebenberg, R.W. Armstrong and J.J. Gilman.Materials Research Society, Pittsburgh, PA. 1993. p. 367.

"Infrared Fiber Optic Diagnostic for Solid Propellant Combustion." J. Wormhoudt and P.L.Kebabian. 29th JANNAF Combustion Proceedings, CPIA Publication 593, Vol. 2. 1992, p. 263.

"Infrared Fiber Optic Diagnostic Observations of Solid Propellant Combustion." J. Wormhoudtand P.L. Kebabian. and C.E. Kolb, to be submitted to Combustion and Flame.

3(9

M0. LIST OF PARTICIPATING SCIENTIFIC PERSONNEL

Dr. Charles E. Kolb was the Principal Investigator for the project. Dr. Joda Wormhoudt

supervised the observations and reduced the data, and Dr. Paul Kebabian assisted with experiment

design, construction, and testing.

Mr. Warren E. Goodwin constructed all the apparatus and test articles, and assisted in the

observations.

40

11. REPORT OF INVENTIONS

This project did not result in any inventions, and we do not anticipate any patent application

based on contract work.

41